net/failsafe: add plug-in support
authorGaetan Rivet <gaetan.rivet@6wind.com>
Tue, 18 Jul 2017 12:48:15 +0000 (14:48 +0200)
committerFerruh Yigit <ferruh.yigit@intel.com>
Wed, 19 Jul 2017 13:25:42 +0000 (16:25 +0300)
Periodically check for the existence of a device.
If a device has not been initialized and exists on the system, then it
is probed and configured.

The configuration process strives to synchronize the states between the
plugged-in sub-device and the fail-safe device.

Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
Acked-by: Olga Shern <olgas@mellanox.com>
doc/guides/nics/fail_safe.rst
drivers/net/failsafe/Makefile
drivers/net/failsafe/failsafe.c
drivers/net/failsafe/failsafe_args.c
drivers/net/failsafe/failsafe_eal.c
drivers/net/failsafe/failsafe_ether.c [new file with mode: 0644]
drivers/net/failsafe/failsafe_ops.c
drivers/net/failsafe/failsafe_private.h

index c20696e..0005892 100644 (file)
@@ -98,6 +98,11 @@ Fail-safe command line parameters
   address is generated, that will be subsequently applied to all sub-device once
   they are probed.
 
   address is generated, that will be subsequently applied to all sub-device once
   they are probed.
 
+- **hotplug_poll** parameter [UINT64] (default **2000**)
+
+  This parameter allows the user to configure the amount of time in milliseconds
+  between two slave upkeep round.
+
 Usage example
 ~~~~~~~~~~~~~
 
 Usage example
 ~~~~~~~~~~~~~
 
@@ -140,6 +145,20 @@ access, and in particular, using the ``RTE_ETH_FOREACH_DEV`` macro to iterate
 over ethernet devices, instead of directly accessing them or by writing one's
 own device iterator.
 
 over ethernet devices, instead of directly accessing them or by writing one's
 own device iterator.
 
+Plug-in feature
+---------------
+
+A sub-device can be defined without existing on the system when the fail-safe
+PMD is initialized. Upon probing this device, the fail-safe PMD will detect its
+absence and postpone its use. It will then register for a periodic check on any
+missing sub-device.
+
+During this time, the fail-safe PMD can be used normally, configured and told to
+emit and receive packets. It will store any applied configuration, and try to
+apply it upon the probing of its missing sub-device. After this configuration
+pass, the new sub-device will be synchronized with other sub-devices, i.e. be
+started if the fail-safe PMD has been started by the user before.
+
 Fail-safe glossary
 ------------------
 
 Fail-safe glossary
 ------------------
 
@@ -153,6 +172,10 @@ Preferred device : Primary device
     It is the main sub-device and is used as target for configuration
     operations if there is any ambiguity.
 
     It is the main sub-device and is used as target for configuration
     operations if there is any ambiguity.
 
+Upkeep round
+    Periodical process when slaves are serviced. Each devices having a state
+    different to that of the fail-safe device itself, is synchronized with it.
+
 Slave
     In the context of the fail-safe PMD, synonymous to sub-device.
 
 Slave
     In the context of the fail-safe PMD, synonymous to sub-device.
 
index e87732a..9bc7409 100644 (file)
@@ -44,6 +44,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_args.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_eal.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_ops.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_rxtx.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_eal.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_ops.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_rxtx.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_ether.c
 
 # No exported include files
 
 
 # No exported include files
 
index 806c305..8db6355 100644 (file)
@@ -81,6 +81,72 @@ fs_sub_device_free(struct rte_eth_dev *dev)
        rte_free(PRIV(dev)->subs);
 }
 
        rte_free(PRIV(dev)->subs);
 }
 
+static void fs_hotplug_alarm(void *arg);
+
+int
+failsafe_hotplug_alarm_install(struct rte_eth_dev *dev)
+{
+       int ret;
+
+       if (dev == NULL)
+               return -EINVAL;
+       if (PRIV(dev)->pending_alarm)
+               return 0;
+       ret = rte_eal_alarm_set(hotplug_poll * 1000,
+                               fs_hotplug_alarm,
+                               dev);
+       if (ret) {
+               ERROR("Could not set up plug-in event detection");
+               return ret;
+       }
+       PRIV(dev)->pending_alarm = 1;
+       return 0;
+}
+
+int
+failsafe_hotplug_alarm_cancel(struct rte_eth_dev *dev)
+{
+       int ret = 0;
+
+       if (PRIV(dev)->pending_alarm) {
+               rte_errno = 0;
+               rte_eal_alarm_cancel(fs_hotplug_alarm, dev);
+               if (rte_errno) {
+                       ERROR("rte_eal_alarm_cancel failed (errno: %s)",
+                             strerror(rte_errno));
+                       ret = -rte_errno;
+               } else {
+                       PRIV(dev)->pending_alarm = 0;
+               }
+       }
+       return ret;
+}
+
+static void
+fs_hotplug_alarm(void *arg)
+{
+       struct rte_eth_dev *dev = arg;
+       struct sub_device *sdev;
+       int ret;
+       uint8_t i;
+
+       if (!PRIV(dev)->pending_alarm)
+               return;
+       PRIV(dev)->pending_alarm = 0;
+       FOREACH_SUBDEV(sdev, i, dev)
+               if (sdev->state != PRIV(dev)->state)
+                       break;
+       /* if we have non-probed device */
+       if (i != PRIV(dev)->subs_tail) {
+               ret = failsafe_eth_dev_state_sync(dev);
+               if (ret)
+                       ERROR("Unable to synchronize sub_device state");
+       }
+       ret = failsafe_hotplug_alarm_install(dev);
+       if (ret)
+               ERROR("Unable to set up next alarm");
+}
+
 static int
 fs_eth_dev_create(struct rte_vdev_device *vdev)
 {
 static int
 fs_eth_dev_create(struct rte_vdev_device *vdev)
 {
@@ -126,6 +192,11 @@ fs_eth_dev_create(struct rte_vdev_device *vdev)
        ret = failsafe_eal_init(dev);
        if (ret)
                goto free_args;
        ret = failsafe_eal_init(dev);
        if (ret)
                goto free_args;
+       ret = failsafe_hotplug_alarm_install(dev);
+       if (ret) {
+               ERROR("Could not set up plug-in event detection");
+               goto free_args;
+       }
        mac = &dev->data->mac_addrs[0];
        if (mac_from_arg) {
                /*
        mac = &dev->data->mac_addrs[0];
        if (mac_from_arg) {
                /*
index bde8f2d..733344e 100644 (file)
@@ -32,6 +32,7 @@
  */
 
 #include <string.h>
  */
 
 #include <string.h>
+#include <errno.h>
 
 #include <rte_devargs.h>
 #include <rte_malloc.h>
 
 #include <rte_devargs.h>
 #include <rte_malloc.h>
 typedef int (parse_cb)(struct rte_eth_dev *dev, const char *params,
                uint8_t head);
 
 typedef int (parse_cb)(struct rte_eth_dev *dev, const char *params,
                uint8_t head);
 
+uint64_t hotplug_poll = FAILSAFE_HOTPLUG_DEFAULT_TIMEOUT_MS;
 int mac_from_arg = 0;
 
 const char *pmd_failsafe_init_parameters[] = {
 int mac_from_arg = 0;
 
 const char *pmd_failsafe_init_parameters[] = {
+       PMD_FAILSAFE_HOTPLUG_POLL_KVARG,
        PMD_FAILSAFE_MAC_KVARG,
        NULL,
 };
        PMD_FAILSAFE_MAC_KVARG,
        NULL,
 };
@@ -220,6 +223,24 @@ out:
        return 0;
 }
 
        return 0;
 }
 
+static int
+fs_get_u64_arg(const char *key __rte_unused,
+               const char *value, void *out)
+{
+       uint64_t *u64 = out;
+       char *endptr = NULL;
+
+       if ((value == NULL) || (out == NULL))
+               return -EINVAL;
+       errno = 0;
+       *u64 = strtoull(value, &endptr, 0);
+       if (errno != 0)
+               return -errno;
+       if (endptr == value)
+               return -1;
+       return 0;
+}
+
 static int
 fs_get_mac_addr_arg(const char *key __rte_unused,
                const char *value, void *out)
 static int
 fs_get_mac_addr_arg(const char *key __rte_unused,
                const char *value, void *out)
@@ -271,6 +292,16 @@ failsafe_args_parse(struct rte_eth_dev *dev, const char *params)
                                PMD_FAILSAFE_PARAM_STRING);
                        return -1;
                }
                                PMD_FAILSAFE_PARAM_STRING);
                        return -1;
                }
+               /* PLUG_IN event poll timer */
+               arg_count = rte_kvargs_count(kvlist,
+                               PMD_FAILSAFE_HOTPLUG_POLL_KVARG);
+               if (arg_count == 1) {
+                       ret = rte_kvargs_process(kvlist,
+                                       PMD_FAILSAFE_HOTPLUG_POLL_KVARG,
+                                       &fs_get_u64_arg, &hotplug_poll);
+                       if (ret < 0)
+                               goto free_kvlist;
+               }
                /* MAC addr */
                arg_count = rte_kvargs_count(kvlist,
                                PMD_FAILSAFE_MAC_KVARG);
                /* MAC addr */
                arg_count = rte_kvargs_count(kvlist,
                                PMD_FAILSAFE_MAC_KVARG);
@@ -284,6 +315,7 @@ failsafe_args_parse(struct rte_eth_dev *dev, const char *params)
                        mac_from_arg = 1;
                }
        }
                        mac_from_arg = 1;
                }
        }
+       PRIV(dev)->state = DEV_PARSED;
 free_kvlist:
        rte_kvargs_free(kvlist);
        return ret;
 free_kvlist:
        rte_kvargs_free(kvlist);
        return ret;
index a74edea..8a3836c 100644 (file)
@@ -72,37 +72,14 @@ fs_bus_init(struct rte_eth_dev *dev)
 int
 failsafe_eal_init(struct rte_eth_dev *dev)
 {
 int
 failsafe_eal_init(struct rte_eth_dev *dev)
 {
-       struct sub_device *sdev;
-       uint8_t i;
        int ret;
 
        ret = fs_bus_init(dev);
        if (ret)
                return ret;
        int ret;
 
        ret = fs_bus_init(dev);
        if (ret)
                return ret;
-       /*
-        * We only update TX_SUBDEV if we are not started.
-        * If a sub_device is emitting, we will switch the TX_SUBDEV to the
-        * preferred port only upon starting it, so that the switch is smoother.
-        */
-       if (PREFERRED_SUBDEV(dev)->state >= DEV_PROBED) {
-               if (TX_SUBDEV(dev) != PREFERRED_SUBDEV(dev) &&
-                   (TX_SUBDEV(dev) == NULL ||
-                    (TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_STARTED))) {
-                       DEBUG("Switching tx_dev to preferred sub_device");
-                       PRIV(dev)->subs_tx = 0;
-               }
-       } else {
-               if ((TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_PROBED) ||
-                   TX_SUBDEV(dev) == NULL) {
-                       /* Using first probed device */
-                       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_PROBED) {
-                               DEBUG("Switching tx_dev to sub_device %d",
-                                     i);
-                               PRIV(dev)->subs_tx = i;
-                               break;
-                       }
-               }
-       }
+       if (PRIV(dev)->state < DEV_PROBED)
+               PRIV(dev)->state = DEV_PROBED;
+       fs_switch_dev(dev);
        return 0;
 }
 
        return 0;
 }
 
@@ -134,5 +111,6 @@ failsafe_eal_uninit(struct rte_eth_dev *dev)
        ret = fs_bus_uninit(dev);
        if (ret)
                return ret;
        ret = fs_bus_uninit(dev);
        if (ret)
                return ret;
+       PRIV(dev)->state = DEV_PROBED - 1;
        return 0;
 }
        return 0;
 }
diff --git a/drivers/net/failsafe/failsafe_ether.c b/drivers/net/failsafe/failsafe_ether.c
new file mode 100644 (file)
index 0000000..7910952
--- /dev/null
@@ -0,0 +1,228 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <unistd.h>
+
+#include "failsafe_private.h"
+
+static int
+fs_eth_dev_conf_apply(struct rte_eth_dev *dev,
+               struct sub_device *sdev)
+{
+       struct rte_eth_dev *edev;
+       struct rte_vlan_filter_conf *vfc1;
+       struct rte_vlan_filter_conf *vfc2;
+       uint32_t i;
+       int ret;
+
+       edev = ETH(sdev);
+       /* RX queue setup */
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               struct rxq *rxq;
+
+               rxq = dev->data->rx_queues[i];
+               ret = rte_eth_rx_queue_setup(PORT_ID(sdev), i,
+                               rxq->info.nb_desc, rxq->socket_id,
+                               &rxq->info.conf, rxq->info.mp);
+               if (ret) {
+                       ERROR("rx_queue_setup failed");
+                       return ret;
+               }
+       }
+       /* TX queue setup */
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               struct txq *txq;
+
+               txq = dev->data->tx_queues[i];
+               ret = rte_eth_tx_queue_setup(PORT_ID(sdev), i,
+                               txq->info.nb_desc, txq->socket_id,
+                               &txq->info.conf);
+               if (ret) {
+                       ERROR("tx_queue_setup failed");
+                       return ret;
+               }
+       }
+       /* dev_link.link_status */
+       if (dev->data->dev_link.link_status !=
+           edev->data->dev_link.link_status) {
+               DEBUG("Configuring link_status");
+               if (dev->data->dev_link.link_status)
+                       ret = rte_eth_dev_set_link_up(PORT_ID(sdev));
+               else
+                       ret = rte_eth_dev_set_link_down(PORT_ID(sdev));
+               if (ret) {
+                       ERROR("Failed to apply link_status");
+                       return ret;
+               }
+       } else {
+               DEBUG("link_status already set");
+       }
+       /* promiscuous */
+       if (dev->data->promiscuous != edev->data->promiscuous) {
+               DEBUG("Configuring promiscuous");
+               if (dev->data->promiscuous)
+                       rte_eth_promiscuous_enable(PORT_ID(sdev));
+               else
+                       rte_eth_promiscuous_disable(PORT_ID(sdev));
+       } else {
+               DEBUG("promiscuous already set");
+       }
+       /* all_multicast */
+       if (dev->data->all_multicast != edev->data->all_multicast) {
+               DEBUG("Configuring all_multicast");
+               if (dev->data->all_multicast)
+                       rte_eth_allmulticast_enable(PORT_ID(sdev));
+               else
+                       rte_eth_allmulticast_disable(PORT_ID(sdev));
+       } else {
+               DEBUG("all_multicast already set");
+       }
+       /* MTU */
+       if (dev->data->mtu != edev->data->mtu) {
+               DEBUG("Configuring MTU");
+               ret = rte_eth_dev_set_mtu(PORT_ID(sdev), dev->data->mtu);
+               if (ret) {
+                       ERROR("Failed to apply MTU");
+                       return ret;
+               }
+       } else {
+               DEBUG("MTU already set");
+       }
+       /* default MAC */
+       DEBUG("Configuring default MAC address");
+       ret = rte_eth_dev_default_mac_addr_set(PORT_ID(sdev),
+                       &dev->data->mac_addrs[0]);
+       if (ret) {
+               ERROR("Setting default MAC address failed");
+               return ret;
+       }
+       /* additional MAC */
+       if (PRIV(dev)->nb_mac_addr > 1)
+               DEBUG("Configure additional MAC address%s",
+                       (PRIV(dev)->nb_mac_addr > 2 ? "es" : ""));
+       for (i = 1; i < PRIV(dev)->nb_mac_addr; i++) {
+               struct ether_addr *ea;
+
+               ea = &dev->data->mac_addrs[i];
+               ret = rte_eth_dev_mac_addr_add(PORT_ID(sdev), ea,
+                               PRIV(dev)->mac_addr_pool[i]);
+               if (ret) {
+                       char ea_fmt[ETHER_ADDR_FMT_SIZE];
+
+                       ether_format_addr(ea_fmt, ETHER_ADDR_FMT_SIZE, ea);
+                       ERROR("Adding MAC address %s failed", ea_fmt);
+               }
+       }
+       /* VLAN filter */
+       vfc1 = &dev->data->vlan_filter_conf;
+       vfc2 = &edev->data->vlan_filter_conf;
+       if (memcmp(vfc1, vfc2, sizeof(struct rte_vlan_filter_conf))) {
+               uint64_t vbit;
+               uint64_t ids;
+               size_t i;
+               uint16_t vlan_id;
+
+               DEBUG("Configuring VLAN filter");
+               for (i = 0; i < RTE_DIM(vfc1->ids); i++) {
+                       if (vfc1->ids[i] == 0)
+                               continue;
+                       ids = vfc1->ids[i];
+                       while (ids) {
+                               vlan_id = 64 * i;
+                               /* count trailing zeroes */
+                               vbit = ~ids & (ids - 1);
+                               /* clear least significant bit set */
+                               ids ^= (ids ^ (ids - 1)) ^ vbit;
+                               for (; vbit; vlan_id++)
+                                       vbit >>= 1;
+                               ret = rte_eth_dev_vlan_filter(
+                                       PORT_ID(sdev), vlan_id, 1);
+                               if (ret) {
+                                       ERROR("Failed to apply VLAN filter %hu",
+                                               vlan_id);
+                                       return ret;
+                               }
+                       }
+               }
+       } else {
+               DEBUG("VLAN filter already set");
+       }
+       return 0;
+}
+
+int
+failsafe_eth_dev_state_sync(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint32_t inactive;
+       int ret;
+       uint8_t i;
+
+       if (PRIV(dev)->state < DEV_PROBED)
+               return 0;
+       ret = failsafe_eal_init(dev);
+       if (ret)
+               return ret;
+       if (PRIV(dev)->state < DEV_ACTIVE)
+               return 0;
+       inactive = 0;
+       FOREACH_SUBDEV(sdev, i, dev)
+               if (sdev->state == DEV_PROBED)
+                       inactive |= UINT32_C(1) << i;
+       ret = dev->dev_ops->dev_configure(dev);
+       if (ret)
+               return ret;
+       FOREACH_SUBDEV(sdev, i, dev) {
+               if (inactive & (UINT32_C(1) << i)) {
+                       ret = fs_eth_dev_conf_apply(dev, sdev);
+                       if (ret) {
+                               ERROR("Could not apply configuration to sub_device %d",
+                                     i);
+                               /* TODO: disable device */
+                               return ret;
+                       }
+               }
+       }
+       /*
+        * If new devices have been configured, check if
+        * the link state has changed.
+        */
+       if (inactive)
+               dev->dev_ops->link_update(dev, 1);
+       if (PRIV(dev)->state < DEV_STARTED)
+               return 0;
+       ret = dev->dev_ops->dev_start(dev);
+       if (ret)
+               return ret;
+       return 0;
+}
index 2d9e057..fdef019 100644 (file)
@@ -89,6 +89,8 @@ fs_dev_configure(struct rte_eth_dev *dev)
                }
                sdev->state = DEV_ACTIVE;
        }
                }
                sdev->state = DEV_ACTIVE;
        }
+       if (PRIV(dev)->state < DEV_ACTIVE)
+               PRIV(dev)->state = DEV_ACTIVE;
        return 0;
 }
 
        return 0;
 }
 
@@ -108,21 +110,9 @@ fs_dev_start(struct rte_eth_dev *dev)
                        return ret;
                sdev->state = DEV_STARTED;
        }
                        return ret;
                sdev->state = DEV_STARTED;
        }
-       if (PREFERRED_SUBDEV(dev)->state == DEV_STARTED) {
-               if (TX_SUBDEV(dev) != PREFERRED_SUBDEV(dev)) {
-                       DEBUG("Switching tx_dev to preferred sub_device");
-                       PRIV(dev)->subs_tx = 0;
-               }
-       } else {
-               if ((TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_STARTED) ||
-                   TX_SUBDEV(dev) == NULL) {
-                       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_STARTED) {
-                               DEBUG("Switching tx_dev to sub_device %d", i);
-                               PRIV(dev)->subs_tx = i;
-                               break;
-                       }
-               }
-       }
+       if (PRIV(dev)->state < DEV_STARTED)
+               PRIV(dev)->state = DEV_STARTED;
+       fs_switch_dev(dev);
        return 0;
 }
 
        return 0;
 }
 
@@ -132,6 +122,7 @@ fs_dev_stop(struct rte_eth_dev *dev)
        struct sub_device *sdev;
        uint8_t i;
 
        struct sub_device *sdev;
        uint8_t i;
 
+       PRIV(dev)->state = DEV_STARTED - 1;
        FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_STARTED) {
                rte_eth_dev_stop(PORT_ID(sdev));
                sdev->state = DEV_STARTED - 1;
        FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_STARTED) {
                rte_eth_dev_stop(PORT_ID(sdev));
                sdev->state = DEV_STARTED - 1;
@@ -183,6 +174,10 @@ fs_dev_close(struct rte_eth_dev *dev)
        struct sub_device *sdev;
        uint8_t i;
 
        struct sub_device *sdev;
        uint8_t i;
 
+       failsafe_hotplug_alarm_cancel(dev);
+       if (PRIV(dev)->state == DEV_STARTED)
+               dev->dev_ops->dev_stop(dev);
+       PRIV(dev)->state = DEV_ACTIVE - 1;
        FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
                DEBUG("Closing sub_device %d", i);
                rte_eth_dev_close(PORT_ID(sdev));
        FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
                DEBUG("Closing sub_device %d", i);
                rte_eth_dev_close(PORT_ID(sdev));
index 62b5e24..a46d1ea 100644 (file)
 #define FAILSAFE_DRIVER_NAME "Fail-safe PMD"
 
 #define PMD_FAILSAFE_MAC_KVARG "mac"
 #define FAILSAFE_DRIVER_NAME "Fail-safe PMD"
 
 #define PMD_FAILSAFE_MAC_KVARG "mac"
+#define PMD_FAILSAFE_HOTPLUG_POLL_KVARG "hotplug_poll"
 #define PMD_FAILSAFE_PARAM_STRING      \
        "dev(<ifc>),"                   \
 #define PMD_FAILSAFE_PARAM_STRING      \
        "dev(<ifc>),"                   \
-       "mac=mac_addr"                  \
+       "mac=mac_addr,"                 \
+       "hotplug_poll=u64"              \
        ""
 
        ""
 
+#define FAILSAFE_HOTPLUG_DEFAULT_TIMEOUT_MS 2000
+
 #define FAILSAFE_MAX_ETHPORTS 2
 #define FAILSAFE_MAX_ETHADDR 128
 
 #define FAILSAFE_MAX_ETHPORTS 2
 #define FAILSAFE_MAX_ETHADDR 128
 
@@ -103,8 +107,22 @@ struct fs_priv {
        uint32_t mac_addr_pool[FAILSAFE_MAX_ETHADDR];
        /* current capabilities */
        struct rte_eth_dev_info infos;
        uint32_t mac_addr_pool[FAILSAFE_MAX_ETHADDR];
        /* current capabilities */
        struct rte_eth_dev_info infos;
+       /*
+        * Fail-safe state machine.
+        * This level will be tracking state of the EAL and eth
+        * layer at large as defined by the user application.
+        * It will then steer the sub_devices toward the same
+        * synchronized state.
+        */
+       enum dev_state state;
+       unsigned int pending_alarm:1; /* An alarm is pending */
 };
 
 };
 
+/* MISC */
+
+int failsafe_hotplug_alarm_install(struct rte_eth_dev *dev);
+int failsafe_hotplug_alarm_cancel(struct rte_eth_dev *dev);
+
 /* RX / TX */
 
 uint16_t failsafe_rx_burst(void *rxq,
 /* RX / TX */
 
 uint16_t failsafe_rx_burst(void *rxq,
@@ -123,10 +141,15 @@ int failsafe_args_count_subdevice(struct rte_eth_dev *dev, const char *params);
 int failsafe_eal_init(struct rte_eth_dev *dev);
 int failsafe_eal_uninit(struct rte_eth_dev *dev);
 
 int failsafe_eal_init(struct rte_eth_dev *dev);
 int failsafe_eal_uninit(struct rte_eth_dev *dev);
 
+/* ETH_DEV */
+
+int failsafe_eth_dev_state_sync(struct rte_eth_dev *dev);
+
 /* GLOBALS */
 
 extern const char pmd_failsafe_driver_name[];
 extern const struct eth_dev_ops failsafe_ops;
 /* GLOBALS */
 
 extern const char pmd_failsafe_driver_name[];
 extern const struct eth_dev_ops failsafe_ops;
+extern uint64_t hotplug_poll;
 extern int mac_from_arg;
 
 /* HELPERS */
 extern int mac_from_arg;
 
 /* HELPERS */
@@ -205,4 +228,39 @@ fs_find_next(struct rte_eth_dev *dev, uint8_t sid,
        return sid;
 }
 
        return sid;
 }
 
+static inline void
+fs_switch_dev(struct rte_eth_dev *dev)
+{
+       enum dev_state req_state;
+
+       req_state = PRIV(dev)->state;
+       if (PREFERRED_SUBDEV(dev)->state >= req_state) {
+               if (TX_SUBDEV(dev) != PREFERRED_SUBDEV(dev) &&
+                   (TX_SUBDEV(dev) == NULL ||
+                    (req_state == DEV_STARTED) ||
+                    (TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_STARTED))) {
+                       DEBUG("Switching tx_dev to preferred sub_device");
+                       PRIV(dev)->subs_tx = 0;
+               }
+       } else if ((TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < req_state) ||
+                  TX_SUBDEV(dev) == NULL) {
+               struct sub_device *sdev;
+               uint8_t i;
+
+               /* Using acceptable device */
+               FOREACH_SUBDEV_STATE(sdev, i, dev, req_state) {
+                       DEBUG("Switching tx_dev to sub_device %d",
+                             i);
+                       PRIV(dev)->subs_tx = i;
+                       break;
+               }
+       } else if (TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < req_state) {
+               DEBUG("No device ready, deactivating tx_dev");
+               PRIV(dev)->subs_tx = PRIV(dev)->subs_tail;
+       } else {
+               return;
+       }
+       rte_wmb();
+}
+
 #endif /* _RTE_ETH_FAILSAFE_PRIVATE_H_ */
 #endif /* _RTE_ETH_FAILSAFE_PRIVATE_H_ */